feat: Add user-managed genre preferences with catalog picker

[ericdjm]

Implement comprehensive genre management system allowing users to customize their top genres beyond Spotify's auto-detected preferences. This enables better control over generation quality and personalized taste profiles.

Backend Changes:

• Add Genre and UserGenre models with normalized name indexing
• Create database migration (f3a4b5c6d7e8) with seed genres catalog
• Implement genre catalog service with seeding and get-or-create helpers
• Add genre management routes (GET catalog, POST add, DELETE remove)
• Update Spotify profile endpoint to load and persist user genres
• Refactor taste_analyzer to extract reusable helper functions:
  • compute_avg_popularity: Calculate average artist popularity
  • derive_mood_tags: Generate mood tags from genres and popularity
  • generate_summary: Create taste summary sentence
• Update taste_analyzer logic to use user-managed genres when available
• Add new schemas for genre catalog and user genre operations
• Enforce 20 genre maximum limit per user

Database Schema:

• genres table: id, name, normalized_name (unique), created_at
• user_genres table: id, user_id, genre_id (composite unique), created_at
• Unique constraints to prevent duplicates
• Cascade delete on user and genre removal
• Indexes on foreign keys for query performance

Frontend Changes:

• Add genre picker UI to TasteDisplay component
• Implement add/remove genre functionality with IconButton controls
• Add genre catalog dropdown with search/filter
• Display genre removal buttons on each tag
• Add loading states for catalog and genre actions
• Implement optimistic UI updates with profile refresh
• Add toast notifications for errors and limits
• Add API methods: getGenreCatalog, addUserGenre, deleteUserGenre
• Wire up genre updates to profile state management in App.tsx

User Experience Improvements:

• Users can now curate their top genres list manually
• Genre changes automatically refresh taste profile analysis
• Mood tags and summary sentence update based on selected genres
• Visual feedback for all genre operations
• Clear messaging when hitting the 20 genre limit
• Persisted preferences across sessions

Technical Details:

• Genre normalization ensures consistent matching (lowercase, trimmed)
• Seed catalog includes 20 starter genres (ambient, art pop, etc.)
• Auto-seed user genres from Spotify on first profile load
• Parallel genre catalog loading to avoid blocking profile display
• Proper error handling with rollback on integrity violations
• Type-safe operations with Pydantic schemas and TypeScript interfaces

Documentation:

• Add generation.md with comprehensive flow documentation
• Document input concept, lyrics topic, and full generation flows
• Explain single-step vs two-step generation paths
• Detail auto-save behavior and instrumental short-circuit logic